We assessed the risks and benefits of the administration of fentanyl during spinal anesthesia in the elderly. Forty patients (70-83 yr) undergoing knee or hip replacement were studied. Preoperatively, cognitive function (minimental state examination [MMSE]), associated pathology, medications, and treatment were evaluated. Patients had spinal anesthesia with 12.5 mg bupivacaine plus saline (SS; n = 21) or 25 mu g fentanyl (FN; n = 19). The number of ailments and drugs per patient were 2.5 and 2.3, respectively; 35%-44% of disorders were untreated, 16%-26% were symptomatic, and 33% were adequately treated. Groups were comparable regarding demographic data and characteristics of the spinal block. Group FN had more pruritus (P < 0.02) and lower Sao(2) (P < 0.007), but prevalence of side effects was similar. Pain intensity (visual analog scale [VAS], facial expression test [FET]) at the time of analgesia request (TAR) was lower in Group FN (P < 0.01). A poor correlation between VAS and FET (range 0.42-0.58) was obtained. MMSE at hospital discharge was no different from preoperative values. Our results show that 25 mu g of spinal fentanyl do not modify spinal anesthesia in the elderly, but induces pruritus and O-2 desaturation. The decrease in postoperative pain intensity and the preservation of cognitive function would justify the use of spinal fentanyl in the elderly.